Synth reporting
Andrew Baxter
10/09/2019
Setting up data
Data from Human Mortality Database and Human Fertility Database was used to calculate birth rates. This was tidied to give Under-18 rates:
| Code | Country | Year | rate | GDPperCap | MF_ratio | MobilePhones | UrbanPop |
|---|---|---|---|---|---|---|---|
| 1 | Austria | 1985 | 8.63 | 9172.10 | 1.04 | 0.13 | 64.31 |
| 1 | Austria | 1986 | 8.44 | 13083.07 | 1.04 | 0.25 | 64.04 |
| 1 | Austria | 1987 | 8.19 | 16392.77 | 1.03 | 0.34 | 63.77 |
| 1 | Austria | 1988 | 7.89 | 17578.62 | 1.04 | 0.48 | 63.50 |
| 1 | Austria | 1989 | 8.58 | 17468.95 | 1.04 | 0.66 | 63.23 |
| 1 | Austria | 1990 | 8.80 | 21680.99 | 1.04 | 0.95 | 62.96 |
Abortion estimates were added to give Under-20 rates:
| Code | Country | Year | pRate | rate | GDPperCap | MF_ratio | MobilePhones | UrbanPop |
|---|---|---|---|---|---|---|---|---|
| 2 | Czechia | 1990 | 71.17 | 46.01 | 3917.16 | 1.05 | 0.00 | 75.22 |
| 2 | Czechia | 1991 | 72.91 | 47.52 | 2878.72 | 1.05 | 0.01 | 75.16 |
| 2 | Czechia | 1992 | 69.70 | 45.14 | 3352.03 | 1.05 | 0.04 | 75.03 |
| 2 | Czechia | 1993 | 62.31 | 42.91 | 3931.74 | 1.05 | 0.14 | 74.90 |
| 2 | Czechia | 1994 | 46.97 | 32.33 | 4601.95 | 1.04 | 0.29 | 74.77 |
| 2 | Czechia | 1995 | 36.53 | 24.43 | 5788.15 | 1.04 | 0.47 | 74.64 |
Iterating through year combinations
For each comparison, I iterated through all combinations of years as special predictors to minimise MSPE (whilst prioritising fewest groupings). For example, for the under-18 basic model with years as special predictors:
it_u18_rateSp <- testSynthIterations(
yrs = 1985:1998,
pred = "rate",
data = synthData_u18[,1:4],
ccodes = u_18_ccodes,
n = 4,
predictors = NULL,
time.optimise = 1985:1998
) %>%
arrange(groups, mspe)Generating Synthetic Control models for all Under-18 comparisons
Model 1: Rate only as predictor
England vs Synthetic Control
Weights
| Country | Weight |
|---|---|
| Lithuania | 0.648 |
| Italy | 0.179 |
| Norway | 0.105 |
| United States of America | 0.056 |
| Poland | 0.005 |
| Czechia | 0.001 |
| Estonia | 0.001 |
| Finland | 0.001 |
| Germany | 0.001 |
| Portugal | 0.001 |
| Switzerland | 0.001 |
| Austria | 0.000 |
| Denmark | 0.000 |
| France | 0.000 |
| Hungary | 0.000 |
| Iceland | 0.000 |
| Netherlands | 0.000 |
| Northern Ireland | 0.000 |
| Scotland | 0.000 |
| Slovenia | 0.000 |
| Spain | 0.000 |
| Sweden | 0.000 |
Placebo testing by country and time
Model 2: GDP as predictor
England vs Synthetic Control
Weights
| Country | Weight |
|---|---|
| Norway | 0.502 |
| United States of America | 0.325 |
| Sweden | 0.173 |
| Austria | 0.000 |
| Denmark | 0.000 |
| Finland | 0.000 |
| France | 0.000 |
| Germany | 0.000 |
| Iceland | 0.000 |
| Italy | 0.000 |
| Netherlands | 0.000 |
| Portugal | 0.000 |
| Spain | 0.000 |
| Switzerland | 0.000 |
Placebo testing by country and time
Model 3: All predictors
England vs Synthetic Control
Weights
| Country | Weight |
|---|---|
| Norway | 0.482 |
| United States of America | 0.322 |
| Sweden | 0.177 |
| Finland | 0.019 |
| Austria | 0.000 |
| Denmark | 0.000 |
| France | 0.000 |
| Germany | 0.000 |
| Iceland | 0.000 |
| Italy | 0.000 |
| Netherlands | 0.000 |
| Portugal | 0.000 |
| Spain | 0.000 |
| Switzerland | 0.000 |
Placebo testing by country and time
Generating Synthetic Control models for all Under-20 comparisons
Model 1: England vs Synthetic Control
England vs Synthetic Control
Weights
| Country | Weight |
|---|---|
| Italy | 0.723 |
| United States of America | 0.211 |
| Netherlands | 0.066 |
| Czechia | 0.000 |
| Denmark | 0.000 |
| Estonia | 0.000 |
| Finland | 0.000 |
| France | 0.000 |
| Germany | 0.000 |
| Hungary | 0.000 |
| Iceland | 0.000 |
| Lithuania | 0.000 |
| New Zealand | 0.000 |
| Norway | 0.000 |
| Poland | 0.000 |
| Portugal | 0.000 |
| Scotland | 0.000 |
| Slovenia | 0.000 |
| Spain | 0.000 |
| Sweden | 0.000 |
| Switzerland | 0.000 |
Placebo testing by country and time
Model 2: GDP as predictor
England vs Synthetic Control
Weights
| Country | Weight |
|---|---|
| Netherlands | 0.581 |
| United States of America | 0.419 |
| Czechia | 0.000 |
| Denmark | 0.000 |
| Finland | 0.000 |
| France | 0.000 |
| Germany | 0.000 |
| Hungary | 0.000 |
| Iceland | 0.000 |
| Italy | 0.000 |
| New Zealand | 0.000 |
| Norway | 0.000 |
| Poland | 0.000 |
| Portugal | 0.000 |
| Spain | 0.000 |
| Sweden | 0.000 |
| Switzerland | 0.000 |
Placebo testing by country and time
Model 3: All predictors
England vs Synthetic Control
Weights
| Country | Weight |
|---|---|
| Netherlands | 0.581 |
| United States of America | 0.419 |
| Czechia | 0.000 |
| Denmark | 0.000 |
| Finland | 0.000 |
| France | 0.000 |
| Germany | 0.000 |
| Hungary | 0.000 |
| Iceland | 0.000 |
| Italy | 0.000 |
| Norway | 0.000 |
| Portugal | 0.000 |
| Spain | 0.000 |
| Sweden | 0.000 |
| Switzerland | 0.000 |